Automatic design of boolean networks for cell differentiation

Cell differentiation is the process that denotes a cell type change, typically from a less specialised type to a more specialised one. Recently, a cell differentiation model based on Boolean networks subject to noise has been proposed. This model reproduces the main abstract properties of cell differentiation, such as the attainment of different degrees of differentiation, deterministic and stochastic differentiation, reversibility, induced pluripotency and cell type change. The generic abstract properties of the model have been already shown to match those of the real biological phenomenon. A direct comparison with specific cell differentiation processes and the identification of genetic network features that are linked to specific differentiation traits requires the design of a suitable Boolean network such that its dynamics matches a set of target properties. To the best of our knowledge, the only current method for addressing this problem is a random generate and test procedure. In this work we present an automatic design method for this purpose, based on metaheuristic algorithms. We devised two variants of the method and tested them against random search on typical abstract differentiation trees. Results, although preliminary, show that our technique is far more efficient than both random search and complete enumeration and it is able to find solutions to instances which were not solved by those techniques.